Chiew, Yee-Meng; Narasimhan, Nandakumar; Chu, Jian Effect ...
Satyadev Nandakumar Santosh Kumar Vangepalli Indian ...PDF/FSDNLN.pdf · 1 / 21 Normality and...
Transcript of Satyadev Nandakumar Santosh Kumar Vangepalli Indian ...PDF/FSDNLN.pdf · 1 / 21 Normality and...
1 / 21
Normality and Finite State Dimension of Liouville Numbers
Satyadev Nandakumar Santosh Kumar VangepalliIndian Institute of Technology Kanpur
September 2013
Introduction
⊲ Introduction
Definition
Examples
Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
2 / 21
Normal Numbers
Introduction
⊲ Definition
Examples
Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
3 / 21
Definition. [Normal Binary Sequence]
Normal Numbers
Introduction
⊲ Definition
Examples
Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
3 / 21
Definition. [Normal Binary Sequence]
An infinite binary sequence ω is said to be normal in base 2 if forevery natural number k and every k-long string w,
limn→∞
|{i ∈ N | ω[i . . . i+ n− 1] = w}|n
=1
2k.
Normal Numbers
Introduction
⊲ Definition
Examples
Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
3 / 21
Definition. [Normal Binary Sequence]
An infinite binary sequence ω is said to be normal in base 2 if forevery natural number k and every k-long string w,
limn→∞
|{i ∈ N | ω[i . . . i+ n− 1] = w}|n
=1
2k.
For example, the asymptotic density of 0s is 1/2,
Normal Numbers
Introduction
⊲ Definition
Examples
Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
3 / 21
Definition. [Normal Binary Sequence]
An infinite binary sequence ω is said to be normal in base 2 if forevery natural number k and every k-long string w,
limn→∞
|{i ∈ N | ω[i . . . i+ n− 1] = w}|n
=1
2k.
For example, the asymptotic density of 0s is 1/2, the asymptoticdensity of 01 is 1
4 , and so on.
Normal Numbers
Introduction
⊲ Definition
Examples
Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
3 / 21
Definition. [Normal Binary Sequence]
An infinite binary sequence ω is said to be normal in base 2 if forevery natural number k and every k-long string w,
limn→∞
|{i ∈ N | ω[i . . . i+ n− 1] = w}|n
=1
2k.
For example, the asymptotic density of 0s is 1/2, the asymptoticdensity of 01 is 1
4 , and so on.
A real number in [0,1] will be called normal if its binaryexpansion is a normal sequence.
Examples of Normal Numbers
Introduction
Definition
⊲ Examples
Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
4 / 21
Are there normal numbers in [0,1]?
Examples of Normal Numbers
Introduction
Definition
⊲ Examples
Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
4 / 21
Are there normal numbers in [0,1]?
Theorem (Borel, 1909). The set of normal numbers in [0,1] hasLebesgue measure 1.
Examples of Normal Numbers
Introduction
Definition
⊲ Examples
Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
4 / 21
Are there normal numbers in [0,1]?
Theorem (Borel, 1909). The set of normal numbers in [0,1] hasLebesgue measure 1.
But are there any constructions known?
Examples of Normal Numbers
Introduction
Definition
⊲ Examples
Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
4 / 21
Are there normal numbers in [0,1]?
Theorem (Borel, 1909). The set of normal numbers in [0,1] hasLebesgue measure 1.
But are there any constructions known?
A simple example of a normal binary sequence:
Examples of Normal Numbers
Introduction
Definition
⊲ Examples
Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
4 / 21
Are there normal numbers in [0,1]?
Theorem (Borel, 1909). The set of normal numbers in [0,1] hasLebesgue measure 1.
But are there any constructions known?
A simple example of a normal binary sequence:
0 1
Examples of Normal Numbers
Introduction
Definition
⊲ Examples
Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
4 / 21
Are there normal numbers in [0,1]?
Theorem (Borel, 1909). The set of normal numbers in [0,1] hasLebesgue measure 1.
But are there any constructions known?
A simple example of a normal binary sequence:
0 1 00 01 10 11
Examples of Normal Numbers
Introduction
Definition
⊲ Examples
Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
4 / 21
Are there normal numbers in [0,1]?
Theorem (Borel, 1909). The set of normal numbers in [0,1] hasLebesgue measure 1.
But are there any constructions known?
A simple example of a normal binary sequence:
0 1 00 01 10 11 000 . . .
Examples of Normal Numbers
Introduction
Definition
⊲ Examples
Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
4 / 21
Are there normal numbers in [0,1]?
Theorem (Borel, 1909). The set of normal numbers in [0,1] hasLebesgue measure 1.
But are there any constructions known?
A simple example of a normal binary sequence:
0 1 00 01 10 11 000 . . .
This is known as the Champernowne sequence (1933).
Properties of Normal Numbers
Introduction
Definition
Examples
⊲ Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
5 / 21
No normal number can be rational.
Properties of Normal Numbers
Introduction
Definition
Examples
⊲ Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
5 / 21
No normal number can be rational.
Definition. A real number is algebraic if it is the root of apolynomial with integer coefficients. e.g.
√2 is the root of the
polynomial x2 − 2.
Properties of Normal Numbers
Introduction
Definition
Examples
⊲ Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
5 / 21
No normal number can be rational.
Definition. A real number is algebraic if it is the root of apolynomial with integer coefficients. e.g.
√2 is the root of the
polynomial x2 − 2.Non-algebraic irrational numbers are called transcendental. (e.g.π, e etc.)
Properties of Normal Numbers
Introduction
Definition
Examples
⊲ Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
5 / 21
No normal number can be rational.
Definition. A real number is algebraic if it is the root of apolynomial with integer coefficients. e.g.
√2 is the root of the
polynomial x2 − 2.Non-algebraic irrational numbers are called transcendental. (e.g.π, e etc.)
Are normal numbers always transcendental? (We do not know)
Properties of Normal Numbers
Introduction
Definition
Examples
⊲ Properties
Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
5 / 21
No normal number can be rational.
Definition. A real number is algebraic if it is the root of apolynomial with integer coefficients. e.g.
√2 is the root of the
polynomial x2 − 2.Non-algebraic irrational numbers are called transcendental. (e.g.π, e etc.)
Are normal numbers always transcendental? (We do not know)
Can we construct normal numbers which are provablytranscendental? What can we say about normal numbers inspecial classes of transcendental numbers?
Transcendence and Normality
Introduction
Definition
Examples
Properties
⊲ Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
6 / 21
Champernowne number is a transcendental number. (Mahler,1937)
Transcendence and Normality
Introduction
Definition
Examples
Properties
⊲ Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
6 / 21
Champernowne number is a transcendental number. (Mahler,1937)
Some transcendental numbers are far from normal.
Example. Liouville constant1 defined as
α =∞∑
i=0
1
2i!.
is a transcendental number.
Transcendence and Normality
Introduction
Definition
Examples
Properties
⊲ Transcendence
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
6 / 21
Champernowne number is a transcendental number. (Mahler,1937)
Some transcendental numbers are far from normal.
Example. Liouville constant1 defined as
α =∞∑
i=0
1
2i!.
is a transcendental number.
α cannot be normal - e.g. 111 never appears in α.
1in base 2
Normality and Diophantine Approximation
Introduction
⊲
Normality andDiophantineApproximation
Liouville Numbers
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
7 / 21
Liouville Numbers
8 / 21
Theorem (Liouville, 1854). A real number γ is an algebraic number of degreed. Then there is a constant C such that for every pair of natural numbers aand b,
∣
∣
∣γ − a
b
∣
∣
∣<
C
bd.
Liouville Numbers
8 / 21
Theorem (Liouville, 1854). A real number γ is an algebraic number of degreed. Then there is a constant C such that for every pair of natural numbers aand b,
∣
∣
∣γ − a
b
∣
∣
∣<
C
bd.
We can define the following class of transcendentals.
Liouville Numbers
8 / 21
Theorem (Liouville, 1854). A real number γ is an algebraic number of degreed. Then there is a constant C such that for every pair of natural numbers aand b,
∣
∣
∣γ − a
b
∣
∣
∣<
C
bd.
We can define the following class of transcendentals.
Definition. A real number β is called a Liouville Number if there is a
sequence of rationals(
piqi
)
∞
i=1such that
∣
∣
∣
∣
β − piqi
∣
∣
∣
∣
≤ 1
qii.
Liouville Numbers
8 / 21
Theorem (Liouville, 1854). A real number γ is an algebraic number of degreed. Then there is a constant C such that for every pair of natural numbers aand b,
∣
∣
∣γ − a
b
∣
∣
∣<
C
bd.
We can define the following class of transcendentals.
Definition. A real number β is called a Liouville Number if there is a
sequence of rationals(
piqi
)
∞
i=1such that
∣
∣
∣
∣
β − piqi
∣
∣
∣
∣
≤ 1
qii.
Liouville’s constant is a Liouville number. (first number provedtranscendental).
Normality of Liouville numbers
Introduction
Normality andDiophantineApproximation
⊲Normality ofLiouville numbers
Question
de Bruijn strings
Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
9 / 21
Normality of Liouville Numbers
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
⊲ Question
de Bruijn strings
Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
10 / 21
Liouville numbers have Lebesgue measure 0. (e.g. Oxtoby, 1980)
Normality of Liouville Numbers
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
⊲ Question
de Bruijn strings
Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
10 / 21
Liouville numbers have Lebesgue measure 0. (e.g. Oxtoby, 1980)They even have Hausdorff dimension 0. (e.g. Oxtoby, 1980).Further, they have constructive Hausdorff Dimension 0. (Staiger,2002)
Normality of Liouville Numbers
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
⊲ Question
de Bruijn strings
Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
10 / 21
Liouville numbers have Lebesgue measure 0. (e.g. Oxtoby, 1980)They even have Hausdorff dimension 0. (e.g. Oxtoby, 1980).Further, they have constructive Hausdorff Dimension 0. (Staiger,2002)
No easy measure-theoretic existence proof. (A measure 0 setmay or may not meet a measure 1 set.)
Normality of Liouville Numbers
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
⊲ Question
de Bruijn strings
Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
10 / 21
Liouville numbers have Lebesgue measure 0. (e.g. Oxtoby, 1980)They even have Hausdorff dimension 0. (e.g. Oxtoby, 1980).Further, they have constructive Hausdorff Dimension 0. (Staiger,2002)
No easy measure-theoretic existence proof. (A measure 0 setmay or may not meet a measure 1 set.)
Liouville’s constant is non-normal.
Normality of Liouville Numbers
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
⊲ Question
de Bruijn strings
Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
10 / 21
Liouville numbers have Lebesgue measure 0. (e.g. Oxtoby, 1980)They even have Hausdorff dimension 0. (e.g. Oxtoby, 1980).Further, they have constructive Hausdorff Dimension 0. (Staiger,2002)
No easy measure-theoretic existence proof. (A measure 0 setmay or may not meet a measure 1 set.)
Liouville’s constant is non-normal.
Surprisingly, there are normal Liouville numbers.
Idea: de Bruijn Strings
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
⊲ de Bruijn strings
Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
11 / 21
A binary string w of length 2n is called a de-Bruijn string oforder n if every n length string occurs exactly once in w, whenlooked in a cyclic manner.
Idea: de Bruijn Strings
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
⊲ de Bruijn strings
Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
11 / 21
A binary string w of length 2n is called a de-Bruijn string oforder n if every n length string occurs exactly once in w, whenlooked in a cyclic manner.
e.g. 0110 is a de-Bruin sequence of order 2:
0110
Idea: de Bruijn Strings
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
⊲ de Bruijn strings
Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
11 / 21
A binary string w of length 2n is called a de-Bruijn string oforder n if every n length string occurs exactly once in w, whenlooked in a cyclic manner.
e.g. 0110 is a de-Bruin sequence of order 2:
0110 0110
Idea: de Bruijn Strings
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
⊲ de Bruijn strings
Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
11 / 21
A binary string w of length 2n is called a de-Bruijn string oforder n if every n length string occurs exactly once in w, whenlooked in a cyclic manner.
e.g. 0110 is a de-Bruin sequence of order 2:
0110 0110 0110
Idea: de Bruijn Strings
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
⊲ de Bruijn strings
Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
11 / 21
A binary string w of length 2n is called a de-Bruijn string oforder n if every n length string occurs exactly once in w, whenlooked in a cyclic manner.
e.g. 0110 is a de-Bruin sequence of order 2:
0110 0110 0110 0110
Idea: de Bruijn Strings
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
⊲ de Bruijn strings
Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
11 / 21
A binary string w of length 2n is called a de-Bruijn string oforder n if every n length string occurs exactly once in w, whenlooked in a cyclic manner.
e.g. 0110 is a de-Bruin sequence of order 2:
0110 0110 0110 0110
Theorem (N. G. de Bruijn 1946, I. J. Good 1946). Let Σ be afinite alphabet, and k ∈ N. Then there exists a de Bruijn stringfrom the alphabet Σ of order k. This string has length |Σ|k.
Idea: de Bruijn Strings
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
⊲ de Bruijn strings
Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
11 / 21
A binary string w of length 2n is called a de-Bruijn string oforder n if every n length string occurs exactly once in w, whenlooked in a cyclic manner.
e.g. 0110 is a de-Bruin sequence of order 2:
0110 0110 0110 0110
Theorem (N. G. de Bruijn 1946, I. J. Good 1946). Let Σ be afinite alphabet, and k ∈ N. Then there exists a de Bruijn stringfrom the alphabet Σ of order k. This string has length |Σ|k.
de Bruijn sequences are a standard tool in the study of normality.
Construction
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
⊲ Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
12 / 21
Denote the lexicographically least binary de Bruijn string of orderk as B(k). Consider the number defined as follows.
Construction
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
⊲ Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
12 / 21
Denote the lexicographically least binary de Bruijn string of orderk as B(k). Consider the number defined as follows.
β =
Construction
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
⊲ Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
12 / 21
Denote the lexicographically least binary de Bruijn string of orderk as B(k). Consider the number defined as follows.
β = B(1)11
Construction
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
⊲ Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
12 / 21
Denote the lexicographically least binary de Bruijn string of orderk as B(k). Consider the number defined as follows.
β = B(1)11
B(2)22
Construction
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
⊲ Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
12 / 21
Denote the lexicographically least binary de Bruijn string of orderk as B(k). Consider the number defined as follows.
β = B(1)11
B(2)22
. . . B(k)kk
. . .
Construction
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
⊲ Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
12 / 21
Denote the lexicographically least binary de Bruijn string of orderk as B(k). Consider the number defined as follows.
β = B(1)11
B(2)22
. . . B(k)kk
. . .
β is a normal Liouville number.
Construction
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
⊲ Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
12 / 21
Denote the lexicographically least binary de Bruijn string of orderk as B(k). Consider the number defined as follows.
β = B(1)11
B(2)22
. . . B(k)kk
. . .
β is a normal Liouville number.
Intuition (Liouville Number): The stages have lengthsapproximately k! as in Liouville’s constant. Hence we can try toform a diophantine approximation as in Liouville’s constant.
Construction
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
⊲ Construction
Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
12 / 21
Denote the lexicographically least binary de Bruijn string of orderk as B(k). Consider the number defined as follows.
β = B(1)11
B(2)22
. . . B(k)kk
. . .
β is a normal Liouville number.
Intuition (Liouville Number): The stages have lengthsapproximately k! as in Liouville’s constant. Hence we can try toform a diophantine approximation as in Liouville’s constant.
Intution (Normality): The basic building block of each stage k isa balanced string of all orders ≤ k.
β is a Liouville number
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
Construction
⊲ Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
13 / 21
The length of stage k is 2k kk.
β is a Liouville number
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
Construction
⊲ Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
13 / 21
The length of stage k is 2k kk. The length of the prefix up toand including stage k is Lk =
∑ki=0 2
iii = O((2k)k).
β is a Liouville number
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
Construction
⊲ Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
13 / 21
The length of stage k is 2k kk. The length of the prefix up toand including stage k is Lk =
∑ki=0 2
iii = O((2k)k).
Consider the rational number bk at level k defined to have thesame binary expansion as β up to stage k, followed by a periodicpattern of B(k + 1).
β is a Liouville number
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
Construction
⊲ Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
13 / 21
The length of stage k is 2k kk. The length of the prefix up toand including stage k is Lk =
∑ki=0 2
iii = O((2k)k).
Consider the rational number bk at level k defined to have thesame binary expansion as β up to stage k, followed by a periodicpattern of B(k + 1). The length of its denominator is O(Lk).
β is a Liouville number
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
Construction
⊲ Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
13 / 21
The length of stage k is 2k kk. The length of the prefix up toand including stage k is Lk =
∑ki=0 2
iii = O((2k)k).
Consider the rational number bk at level k defined to have thesame binary expansion as β up to stage k, followed by a periodicpattern of B(k + 1). The length of its denominator is O(Lk).
|bk − β| ≤ 1
2(2(k+1))k+1≤ 1
(2O((2k)k))k
β is a Liouville number
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
Construction
⊲ Liouvilleness
Normality
Generalization:Finite StateDimension
Open Questions
13 / 21
The length of stage k is 2k kk. The length of the prefix up toand including stage k is Lk =
∑ki=0 2
iii = O((2k)k).
Consider the rational number bk at level k defined to have thesame binary expansion as β up to stage k, followed by a periodicpattern of B(k + 1). The length of its denominator is O(Lk).
|bk − β| ≤ 1
2(2(k+1))k+1≤ 1
(2O((2k)k))k
Hence β is a Liouville number.
β is a normal number
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
Construction
Liouvilleness
⊲ Normality
Generalization:Finite StateDimension
Open Questions
14 / 21
In B(k), every string of length k appears once (wrappingaround).
β is a normal number
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
Construction
Liouvilleness
⊲ Normality
Generalization:Finite StateDimension
Open Questions
14 / 21
In B(k), every string of length k appears once (wrappingaround). Hence every string w of length n ≤ k appears with theright frequency 1
2k−n .
β is a normal number
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
Construction
Liouvilleness
⊲ Normality
Generalization:Finite StateDimension
Open Questions
14 / 21
In B(k), every string of length k appears once (wrappingaround). Hence every string w of length n ≤ k appears with theright frequency 1
2k−n .
Hence for all k ≥ |w|, the frequency of any string w is 12w at the
end of every block B(k).
β is a normal number
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
Construction
Liouvilleness
⊲ Normality
Generalization:Finite StateDimension
Open Questions
14 / 21
In B(k), every string of length k appears once (wrappingaround). Hence every string w of length n ≤ k appears with theright frequency 1
2k−n .
Hence for all k ≥ |w|, the frequency of any string w is 12w at the
end of every block B(k). What about within B(k)?
β is a normal number
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
Construction
Liouvilleness
⊲ Normality
Generalization:Finite StateDimension
Open Questions
14 / 21
In B(k), every string of length k appears once (wrappingaround). Hence every string w of length n ≤ k appears with theright frequency 1
2k−n .
Hence for all k ≥ |w|, the frequency of any string w is 12w at the
end of every block B(k). What about within B(k)?
For every large enough k, length(B(k)) = o(Lk−1).
β is a normal number
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Question
de Bruijn strings
Construction
Liouvilleness
⊲ Normality
Generalization:Finite StateDimension
Open Questions
14 / 21
In B(k), every string of length k appears once (wrappingaround). Hence every string w of length n ≤ k appears with theright frequency 1
2k−n .
Hence for all k ≥ |w|, the frequency of any string w is 12w at the
end of every block B(k). What about within B(k)?
For every large enough k, length(B(k)) = o(Lk−1). Hence thedeviations from balance within B(k) is insignificant for lengthsfrom Lk−1 to Lk−1 + length(Bk).
Generalization: Finite State Dimension
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
⊲
Generalization:Finite StateDimension
Definition
Normality and FSD
Construction
Open Questions
15 / 21
Finite-State Dimension
16 / 21
Consider the n-long prefix of an infinite binary sequence ω. Then for k < n,the frequency of a k-long word w in ω[0 . . . n− 1] is
π(ω[0 . . . n− 1], w) =|{i | 0 ≤ i ≤ n− k − 1 ω[i . . . i+ k − 1] = w}|
n− k − 1.
Finite-State Dimension
16 / 21
Consider the n-long prefix of an infinite binary sequence ω. Then for k < n,the frequency of a k-long word w in ω[0 . . . n− 1] is
π(ω[0 . . . n− 1], w) =|{i | 0 ≤ i ≤ n− k − 1 ω[i . . . i+ k − 1] = w}|
n− k − 1.
The entropy of this probability distribution is
Hk,n(ω) =∑
w∈Σk
−π(ω[0 . . . n− 1], w) log π(ω[0 . . . n− 1], w).
Finite-State Dimension
16 / 21
Consider the n-long prefix of an infinite binary sequence ω. Then for k < n,the frequency of a k-long word w in ω[0 . . . n− 1] is
π(ω[0 . . . n− 1], w) =|{i | 0 ≤ i ≤ n− k − 1 ω[i . . . i+ k − 1] = w}|
n− k − 1.
The entropy of this probability distribution is
Hk,n(ω) =∑
w∈Σk
−π(ω[0 . . . n− 1], w) log π(ω[0 . . . n− 1], w).
The k-entropy rate of ω is defined as
Hk(ω) =1
klim infn→∞
Hk,n(ω).
Finite-State Dimension
16 / 21
Consider the n-long prefix of an infinite binary sequence ω. Then for k < n,the frequency of a k-long word w in ω[0 . . . n− 1] is
π(ω[0 . . . n− 1], w) =|{i | 0 ≤ i ≤ n− k − 1 ω[i . . . i+ k − 1] = w}|
n− k − 1.
The entropy of this probability distribution is
Hk,n(ω) =∑
w∈Σk
−π(ω[0 . . . n− 1], w) log π(ω[0 . . . n− 1], w).
The k-entropy rate of ω is defined as
Hk(ω) =1
klim infn→∞
Hk,n(ω).
The finite-dimensional entropy rate of ω is defined as H(ω) = infk Hk(ω).
Normality and Finite-state dimension
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Definition
⊲Normality andFSD
Construction
Open Questions
17 / 21
A binary sequence is normal if and only if it has finite statedimension 1 (Schnorr and Stimm 1971,
Normality and Finite-state dimension
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Definition
⊲Normality andFSD
Construction
Open Questions
17 / 21
A binary sequence is normal if and only if it has finite statedimension 1 (Schnorr and Stimm 1971, Becher and Heiber 2012.)
Normality and Finite-state dimension
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Definition
⊲Normality andFSD
Construction
Open Questions
17 / 21
A binary sequence is normal if and only if it has finite statedimension 1 (Schnorr and Stimm 1971, Becher and Heiber 2012.)
Every binary sequence has a finite-state dimension between 0and 1.
Normality and Finite-state dimension
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Definition
⊲Normality andFSD
Construction
Open Questions
17 / 21
A binary sequence is normal if and only if it has finite statedimension 1 (Schnorr and Stimm 1971, Becher and Heiber 2012.)
Every binary sequence has a finite-state dimension between 0and 1.
Finite-state dimension thus provides a quantification of the“non-normality” of binary expansions of reals.
Normality and Finite-state dimension
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Definition
⊲Normality andFSD
Construction
Open Questions
17 / 21
A binary sequence is normal if and only if it has finite statedimension 1 (Schnorr and Stimm 1971, Becher and Heiber 2012.)
Every binary sequence has a finite-state dimension between 0and 1.
Finite-state dimension thus provides a quantification of the“non-normality” of binary expansions of reals.
Can we produce Liouville numbers of any given finite-statedimension in [0, 1]?
Liouville numbers of any FSD between 0 and 1
18 / 21
Let m,n be positive integers, m < n. We construct a Liouville number withfinite-state dimension m/n.
αm/n = 0·(
(021
)n−mB(1)m)11 (
(022
)n−mB(2)m)22
. . .(
(02k
)n−mB(k)m)kk
.
Liouville numbers of any FSD between 0 and 1
18 / 21
Let m,n be positive integers, m < n. We construct a Liouville number withfinite-state dimension m/n.
αm/n = 0·(
(021
)n−mB(1)m)11 (
(022
)n−mB(2)m)22
. . .(
(02k
)n−mB(k)m)kk
.
αm/n is a Liouville number with finite-state dimension m/n.
Open Questions
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
⊲ Open Questions
Open Questions
.
19 / 21
Open Questions
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
⊲ Open Questions
.
20 / 21
� Can we construct absolutely normal Liouville numbers? Theyexist. (Bugeaud, 2002)
Open Questions
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
⊲ Open Questions
.
20 / 21
� Can we construct absolutely normal Liouville numbers? Theyexist. (Bugeaud, 2002)
� Can we construct algebraic normal numbers?
.
Introduction
Normality andDiophantineApproximation
Normality ofLiouville numbers
Generalization:Finite StateDimension
Open Questions
Open Questions
⊲ .
21 / 21
Thank You!